基于改进残差网络和动态ReLU肺癌病理图像分类  

Lung cancer pathological image classification based on improved residual network and dynamic ReLU

作　　者：谷宇[1] 李思敏 张宝华[1] 杨立东[1] 吕晓琪[1,2] 张祥松 贾成一[1,4] 贺群 GU Yu;LI Simin;ZHANG Baohua;YANG Lidong;LV Xiaoqi;ZHANG Xiangsong;JIA Chengyi;HE Qun(Key Laboratory of Pattern Recognition and Intelligent Image Processing of Inner Mongolia Autonomous Region,School of Information Engineering,Inner Mongolia University of Science and Technology,Baotou Inner Mongolia 014010,China;School of Information Engineering,Inner Mongolia University of Technology,Hohhot Inner Mongolia 010051,China;BOE Technology Group Corporation Ltd,Beijing 100176,China;China Second Metallurgy Group Corporation Ltd,Baotou Inner Mongolia 014010,China.)

机构地区：[1]内蒙古科技大学信息工程学院,内蒙古自治区模式识别与智能图像处理重点实验室,包头014010 [2]内蒙古工业大学信息工程学院,呼和浩特010051 [3]京东方科技集团股份有限公司,北京100176 [4]中国二冶集团有限公司,包头014010

出　　处：《激光杂志》2023年第5期154-161,共8页Laser Journal

基　　金：国家自然科学基金项目(No.62001255,61962046,62161040,61841204);中央引导地方科技发展资金项目(No.2021ZY0004);内蒙古自治区自然科学基金(No.2019MS06003,2015MS0604);内蒙古自治区科技计划项目(No.2020GG0315,2021GG0082,2021GG0023);草原英才“智能交通大数据分析与应用创新人才”创新人才团队、内蒙古自治区高等学校科学研究项目(No.NJZY145);教育部“春晖计划”合作科研项目(教外司留[2019]1383号)。

摘　　要：为实现肺癌病理图像的准确分类,提出一种基于双重注意力和动态ReLU的残差网络DrANeT,在网络模型中嵌入CBAM双重注意力机制,提高网络提取有用特种图的能力;使用动态ReLU替换网络中所使用的普通ReLU,可以根据不同特征图自动调整参数,提高模型性能。实验结果表明,该网络模型在肺良性图像、肺腺癌图像和肺鳞状细胞癌图像准确率分别达到100.00%、99.96%、99.96%,整个数据集准确率达到99.96%。因此,表明本实验所改进的网络模型可以有效地提高肺癌病理图像分类的准确率。To achieve accurate classification of lung cancer pathological images,a residual network DrANeT based on dual attention and dynamic ReLU is proposed.The CBAM dual attention model is embedded in the network model to improve the ability of the network to extract useful feature map;Dynamic ReLU used to replace ordinary ReLU used in the network can automatically adjust parameters according to different feature maps and improve model performance.The experimental results show that the accuracy of the network model in lung benign images,lung adenocarcinoma images and squamous cell carcinoma of lung images reached 100.00%,99.96%and 99.96%,respectively.Therefore,it shows that the improved network model in this experiment can effectively improve the accuracy of lung cancer pathological image classification.

关 键 词：肺癌病理图像 残差网络 注意力机制 图像分类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]